Google has introduced a new music-generation feature within the Gemini app, leveraging DeepMind’s Lyria 3 model to enable users to create music based on text, images, and videos. The feature, currently in beta, allows users to describe the type of song they want, prompting the app to generate a track complete with lyrics. For example, users can request a “comical R&B slow jam about a sock finding its match,” resulting in a unique 30-second track accompanied by custom cover art created by Nano Banana.
Enhancements in Lyria 3 promise more authentic and intricate music compositions compared to previous models. Users also have the flexibility to adjust elements such as style, vocals, and tempo. Beyond Gemini, Google is extending the Lyria 3 model to YouTube creators through the Dream Track feature, enabling AI-generated music production. This feature was previously limited to U.S.-based creators but is now globally accessible.
Google emphasizes that the purpose of music generation with Lyria 3 is original expression rather than replication of existing artists. When prompted with an artist’s name, Gemini will produce a track with a similar style or mood, ensuring creative inspiration. To maintain authenticity, all content generated by the Lyria 3 model will be marked with a SynthID watermark.
Source: TechCrunch